subimanova(1)                                                    subimanova(1)



NAME
       subimanova - subtracts image averages with analysis of variance

SYNOPSIS
       subimanova

DESCRIPTION
       SUBIMANOVA subtracts one set of average images from another set and
       uses a nested analysis of variance (ANOVA) to find the statistical sig-
       nificance of the difference at each pixel.  It then sets to zero all
       differences less significant than a specified level.  The program can
       output either actual differences or pixel values that reflect the level
       of significance.  In order to do the ANOVA, it must have a standard
       deviation or variance image corresponding to each average image.

       The average and standard deviation/variance images can be ones produced
       by IMAVGSTAT or by other means.  When one starts the program, one des-
       ignates a pair of A files (with average and S.D./VAR images) and a pair
       of B files.  One can then subtract any set of sections in B from any
       set of sections in A; A and B may be the same pair of files.

       The user is responsible for keeping track of how many samples were used
       in making each average, and informing this program of those numbers.
       The program needs these numbers to do the ANOVA.

       Entries to the program:

       Average image file A
       Standard deviation or variance image file A
       Average image file B, or Return if same as file for A
       Standard deviation or variance image file B, or Return if same as
          file for A
       Output image file to store differences in

       0 to use a simple mean when combining the average images of one set,
          or 1 to form a weighted mean, where each average image would be
          weighted by the number of samples combined to form that average.
          In the latter case, the mean would be identical to the average
          image that could be obtain by combining ALL of the samples of
          that set.

       0 if the files have standard deviations in them, or 1 if the files
          have variances

       Number of differences to compute

       For each difference, enter:

          List of section numbers in file A, where ranges are allowed
             (e.g. 0-2,4,7-8).

          List of section numbers in file B, where ranges are allowed

          Number of samples making up those averages for each section in A

          Number of samples making up those averages for each section in B

          Significance level (e.g. 0.05, 0.01, etc).  Differences with less
             than this significance will be set to zero.  Enter a
             negative value to have significant pixels values set to
             the negative of the log of the probability, or to the positive
             log for negative differences.  For example, positive and
             negative differences with a P of 0.01 would be output as
             2 and -2, respectively.


       The infamous Satterthwaite approximation will be used whenever the cri-
       teria for its application are satisfied.

HISTORY
       Written by David Mastronarde,  4/23/90
       4/12/95 changed to use local subroutines instead of NAG ones

BUGS
       Email bug reports to mast at colorado dot edu.



BL3DEMC                              4.7.3                       subimanova(1)